• Proceedings of the Korean Society of Computer Information Conference
• /
• 2021.07a
• /
• pp.735-736
• /
• 2021

인터넷에서 댓글 시스템은 자신의 의사표현을 위한 시스템으로 널리 사용되고 있다. 하지만 이를 악용하여 상대방에 대한 혐오를 드러내기도 한다. 악성댓글에 대한 적절한 대처를 위해 빠르고 정확한 탐지는 필수적이다. 본 연구에서는 악성 댓글 분류 문제를 해결하기 위해서 순서가 있는 분류 레이블의 성질을 활용한 순서형 회귀 (Ordinal regression) 기반의 분류 모델을 제안한다. 일반적인 분류 모형과는 달리 혐오 발언 정도에 따라 다중 레이블을 부여하여 학습을 진행하였다. 실험을 통해 Korean Hate Speech Dataset에 대해 LSTM기반의 모형의 출력층을 다르게 구성하여 순서형 회귀 기반의 모형들의 성능을 비교하였다. 결과적으로 예측 결과에 대한 조율이 가능한 순서형 회귀 모형이 일반적인 순서형 회귀 모형에 비해서 편향된 예측에 대해 추가적인 성능 향상을 보였다.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.67 no.10
• /
• pp.1364-1369
• /
• 2018

Speech emotion recognition, which aims to classify speaker's emotional states through speech signals, is one of the essential tasks for making Human-machine interaction (HMI) more natural and realistic. Voice expressions are one of the main information channels in interpersonal communication. However, existing speech emotion recognition technology has not achieved satisfactory performances, probably because of the lack of effective emotion-related features. This paper provides a survey on various features used for speech emotional recognition and discusses which features or which combinations of the features are valuable and meaningful for the emotional recognition classification. The main aim of this paper is to discuss and compare various approaches used for feature extraction and to propose a basis for extracting useful features in order to improve SER performance.

• Proceedings of the KSPS conference
• /
• 2005.11a
• /
• pp.163-166
• /
• 2005

We propose a speech emotion recognition method for natural human-robot interface. In the proposed method, emotion is classified into 6 classes: Angry, bored, happy, neutral, sad and surprised. Features for an input utterance are extracted from statistics of phonetic and prosodic information. Phonetic information includes log energy, shimmer, formant frequencies, and Teager energy; Prosodic information includes pitch, jitter, duration, and rate of speech. Finally a patten classifier based on Gaussian support vector machines decides the emotion class of the utterance. We record speech commands and dialogs uttered at 2m away from microphones in 5different directions. Experimental results show that the proposed method yields 59% classification accuracy while human classifiers give about 50%accuracy, which confirms that the proposed method achieves performance comparable to a human.

• The Journal of the Acoustical Society of Korea
• /
• v.23 no.3
• /
• pp.253-261
• /
• 2004

In this paper, a new voice personality transformation method is proposed. which modifies speaker-dependent feature variables in the speech signals. The proposed method takes the cepstrum vectors and pitch as the transformation paremeters, which represent vocal tract transfer function and excitation signals, respectively. To transform these parameters, a multiple response classification and regression tree (MR-CART) is employed. MR-CART is the vector extended version of a conventional CART, whose response is given by the vector form. We evaluated the performance of the proposed method by comparing with a previously proposed codebook mapping method. We also quantitatively analyzed the performance of voice transformation and the complexities according to various observations. From the experimental results for 4 speakers, the proposed method objectively outperforms a conventional codebook mapping method. and we also observed that the transformed speech sounds closer to target speech.

• The Journal of the Acoustical Society of Korea
• /
• v.21 no.3
• /
• pp.285-292
• /
• 2002

One of the problems that should be solved in Text-To-Speech (TTS) is discontinuities at unit-joining points. To cope with this problem, a smoothing method using a low-pass filter is employed in this paper, In the proposed soothing method, a filter coefficient that controls the amount of smoothing is determined according to contort information to be synthesized. This method efficiently reduces both discontinuities at unit-joining points and artifacts caused by undesired smoothing. The amount of smoothing is determined with discontinuities around unit-joins points in the current synthesized speech and discontinuities predicted from context. The discontinuity predictor is implemented by CART that has context feature variables. To evaluate the performance of the proposed method, a corpus-based concatenative TTS was used as a baseline system. More than 6075 of listeners realized that the quality of the synthesized speech through the proposed smoothing is superior to that of non-smoothing synthesized speech in both naturalness and intelligibility.

• The Journal of the Acoustical Society of Korea
• /
• v.17 no.3E
• /
• pp.12-21
• /
• 1998

In this paper, we describe the one channel five-way, V/U/M/N/S (Voice/Unvoice/Nasal/Silent), classification algorithm for automatically classifying speech. The decision making process is viewed as a pattern viewed as a pattern recognition problem. Two aspects of the algorithm are developed: feature selection and classifier type. The feature selection procedure is studied for identifying a set of features to make V/U/M/N/S classification. The classifiers used are a vector quantization (VQ), a neural network(NN), and a decision tree method. Actual five sentences spoken by six speakers, three male and three female, are tested with proposed classifiers. From a set of measurement tests, the proposed classifiers show fairly good accuracy for V/U/M/N/S decision.

• Speech Sciences
• /
• v.9 no.4
• /
• pp.265-274
• /
• 2002

In this paper we propose an automatic segmentation system that outputs the time alignment information of phoneme boundary using Viterbi search with HMM (Hidden Markov Model) and corrects these results by an UVS (unvoiced/voiced/silence) classification algorithm. We selecte a set of 39 monophones and a set of 647 extended phones for HMM models. For the UVS classification we use the feature parameters such as ZCR (Zero Crossing Rate), log energy, spectral distribution. The result of forced alignment using the extended phone set is 11% better than that of the monophone set. The UVS classification algorithm shows high performance to correct the segmentation results.

• Journal of Broadcast Engineering
• /
• v.20 no.1
• /
• pp.164-170
• /
• 2015

This paper deals with the adaptation of classification model in the binary mask approach to suppress noise in the noisy environment. The binary mask estimation approach is known to improve speech intelligibility of noisy speech. However, the same type of noisy data for the test data should be included in the training data for building the classification model of binary mask estimation. The eigenvoice adaptation is applied to the noise-independent classification model and the adapted model is used as noise-dependent model. The results are reported in Hit rates and False alarm rates. The experimental results confirmed that the accuracy of classification is improved as the number of adaptation sentences increases.

• ETRI Journal
• /
• v.30 no.6
• /
• pp.865-867
• /
• 2008

A conventional environment adaptation for robust speech recognition is usually conducted using transform-based techniques. Here, we present a discriminative adaptation strategy based on a multi-condition-trained model, and propose a new method to provide universal application to a new environment using the environment's specific conditions. Experimental results show that a speech recognition system adapted using the proposed method works successfully for other conditions as well as for those of the new environment.

• ETRI Journal
• /
• v.36 no.5
• /
• pp.714-720
• /
• 2014

In this paper, we propose the use of data-driven probabilistic utterance-level decision logic to improve Weighted Finite State Transducer (WFST)-based endpoint detection. In general, endpoint detection is dealt with using two cascaded decision processes. The first process is frame-level speech/non-speech classification based on statistical hypothesis testing, and the second process is a heuristic-knowledge-based utterance-level speech boundary decision. To handle these two processes within a unified framework, we propose a WFST-based approach. However, a WFST-based approach has the same limitations as conventional approaches in that the utterance-level decision is based on heuristic knowledge and the decision parameters are tuned sequentially. Therefore, to obtain decision knowledge from a speech corpus and optimize the parameters at the same time, we propose the use of data-driven probabilistic utterance-level decision logic. The proposed method reduces the average detection failure rate by about 14% for various noisy-speech corpora collected for an endpoint detection evaluation.